Post

How does one get an objective set of benchmarks that compare agent harnesses instead of the model? So many bench websites compare the model/harness combo but none compare a bunch of agent harnesses with a single model We simply don’t know which harnesses are the best